Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A method for eye-tracking, comprising: obtaining a set of face grayscale images to be detected, wherein the set of face grayscale images comprises a plurality of frames of face grayscale images; judging whether a contour of an iris is determined in an N-th frame of the plurality of frames, wherein N is a positive integer; when the contour of the iris is not determined in the N-th frame, detecting a pupil in the N-th frame, and determining a central position of the pupil in the N-th frame; determining an area of the N-th frame of the face grayscale images corresponding to an eyeball window, wherein the central position of the pupil is a center of the area; and determining the contour of the iris in the N-th frame according to the area corresponding to the eyeball window.
Eye-tracking systems analyze facial images to determine gaze direction, but existing methods often struggle with accurately detecting iris contours in low-quality or partially occluded frames. This invention addresses the problem by providing a robust method for iris contour detection in grayscale face images, even when initial detection fails. The method processes a sequence of grayscale face images, each containing multiple frames. For each frame, it first checks whether the iris contour can be determined. If not, it detects the pupil and calculates its central position. Using this central position, it defines an eyeball window—a region of interest around the pupil. The system then analyzes this window to accurately determine the iris contour. This approach improves reliability by focusing detection on a localized area, reducing errors caused by noise or partial occlusions. By dynamically adjusting the detection region based on pupil position, the method enhances accuracy in challenging conditions, such as low lighting or rapid eye movements. The technique is particularly useful in applications requiring precise gaze tracking, such as medical diagnostics, virtual reality, or human-computer interaction. The iterative process ensures continuous tracking even when initial frame analysis fails, maintaining system robustness.
2. The method according to claim 1 , wherein when the contour of the iris is determined in the N-th frame, then the method further comprises: making the central position of the pupil determined according to the contour of the iris in the N-th frame as a central position of the pupil in an (N+1)-th frame; determining an area of the (N+1)-th frame of the face grayscale images corresponding to an eyeball window, wherein the central position of the pupil in the (N+1)-th frame is a center of the area of the (N+1)-th frame corresponding to the eyeball window; and determining a contour of the iris in the (N+1)-th frame according to a preset condition and the area of the (N+1)-th frame corresponding to the eyeball window.
This invention relates to iris recognition systems, specifically improving the accuracy and efficiency of iris contour detection in sequential video frames. The problem addressed is the computational cost and potential inaccuracies in repeatedly determining the pupil's central position and iris contour in each frame of a face grayscale image sequence. The solution involves a predictive approach where the pupil's central position from a current frame (N-th frame) is used to estimate the same position in the subsequent frame (N+1-th frame). This reduces redundant calculations by leveraging temporal continuity between frames. The method first identifies the iris contour in the N-th frame and uses its pupil's central position as a reference for the N+1-th frame. An eyeball window area is then defined around this predicted central position in the N+1-th frame. The iris contour in the N+1-th frame is subsequently determined based on a preset condition and the defined eyeball window area. This approach minimizes processing overhead while maintaining detection accuracy by exploiting the gradual movement of the eye between consecutive frames. The preset condition may include thresholds or algorithms for edge detection, contrast analysis, or other image processing techniques to refine the iris contour. The method is particularly useful in real-time iris recognition systems where computational efficiency is critical.
3. The method according to claim 1 , wherein detecting the pupil in the N-th frame, and determining the central position of the pupil in the N-th frame comprises: performing horizontal integral projection for the N-th frame, and obtaining a horizontal projection curve of the N-th frame; determining coordinates of left and right eyeballs in a vertical direction of the N-th frame according to the horizontal projection curve of the N-th frame; performing vertical integral projection for an area of the N-th frame corresponding to coordinates of the left and right eyeballs in the vertical direction of the N-th frame, and obtaining a vertical projection curve of the N-th frame; determining coordinates of the left and right eyeballs in a horizontal direction of the N-th frame according to the vertical projection curve of the N-th frame; and determining the central position of the pupil according to the coordinates of the left and right eyeballs in the vertical and horizontal directions of the N-th frame.
This invention relates to pupil detection in video frames, specifically improving accuracy in determining the central position of a pupil within an image frame. The method addresses challenges in real-time eye-tracking systems where pupil detection must be precise despite variations in lighting, head movement, or occlusion. The process involves analyzing a sequence of video frames, where each frame (referred to as the N-th frame) undergoes horizontal and vertical integral projection to isolate the pupil. First, horizontal integral projection is performed on the N-th frame to generate a horizontal projection curve, which helps identify the vertical coordinates of the left and right eyeballs. Next, vertical integral projection is applied to the region of the frame corresponding to these vertical coordinates, producing a vertical projection curve that determines the horizontal coordinates of the eyeballs. The central position of the pupil is then calculated using these vertical and horizontal coordinates. This approach enhances pupil detection accuracy by leveraging projection-based localization, reducing errors caused by noise or partial occlusions. The method is particularly useful in applications requiring precise gaze tracking, such as medical diagnostics, virtual reality interfaces, or human-computer interaction systems.
4. The method according to claim 3 , wherein performing the horizontal integral projection for the N-th frame and obtaining the horizontal projection curve of the N-th frame comprises: removing pixels having grayscale values above a first threshold in the N-th frame, then performing horizontal integral projection for the N-th frame, and obtaining the horizontal projection curve of the N-th frame.
This invention relates to image processing techniques for analyzing frames in a video sequence, specifically focusing on horizontal integral projection to extract features from video frames. The problem addressed involves accurately obtaining horizontal projection curves from video frames while minimizing noise and irrelevant data that could distort the analysis. The method involves processing a sequence of video frames, where each frame is subjected to a horizontal integral projection to generate a horizontal projection curve. Before performing the projection, the method removes pixels with grayscale values exceeding a first threshold. This step filters out high-intensity pixels, which may represent noise or irrelevant features, ensuring that the projection curve accurately reflects the relevant structures in the frame. The horizontal integral projection is then performed on the filtered frame, resulting in a horizontal projection curve that represents the distribution of pixel intensities across the frame's horizontal axis. This technique is particularly useful in applications such as object detection, motion analysis, and video compression, where accurate feature extraction from video frames is essential. By removing high-intensity noise before projection, the method improves the reliability of subsequent analysis steps. The approach can be applied in various domains, including surveillance, medical imaging, and autonomous systems, where precise frame analysis is critical.
5. The method according to claim 3 , wherein determining the coordinates of the left and right eyeballs in the vertical direction of the N-th frame according to the horizontal projection curve of the N-th frame comprises: preprocessing the horizontal projection curve of the N-th frame, and determining coordinates in the N-th frame corresponding to an area between a second trough and a third trough of the preprocessed horizontal projection curve of the N-th frame as the coordinates of the left and right eyeballs in the vertical direction of the N-th frame; wherein the horizontal projection curve of the N-th frame is preprocessed by selecting values of peaks in the horizontal projection curve, and a distance of each of the peaks along the vertical direction of the N-th frame from a nearest trough or crest of the horizontal projection curve is greater than a second threshold.
This invention relates to a method for determining the vertical coordinates of left and right eyeballs in a video frame using horizontal projection curve analysis. The method addresses the challenge of accurately locating eye positions in images or video frames, which is critical for applications like gaze tracking, facial recognition, and augmented reality. The process involves analyzing the horizontal projection curve of a frame, which represents the distribution of pixel intensities along horizontal lines. The curve is preprocessed by selecting peaks where the distance from the nearest trough or crest exceeds a predefined threshold, ensuring only significant features are considered. The coordinates of the left and right eyeballs in the vertical direction are then determined by identifying the area between the second and third troughs of the preprocessed curve. This region corresponds to the vertical span where the eyes are likely located, providing a robust way to isolate eye positions even in varying lighting or facial orientations. The method leverages projection curve analysis to enhance accuracy and reliability in eye detection, reducing false positives and improving performance in real-time applications. By focusing on key features of the projection curve, it efficiently narrows down the search area for eye localization, making it suitable for integration into computer vision systems.
6. The method according to claim 3 , wherein determining the coordinates of the left and right eyeballs in the horizontal direction of the N-th frame according to the vertical projection curve of the N-th frame comprises: preprocessing the vertical projection curve of the N-th frame, and determining coordinates, in the N-th frame, corresponding to two symmetric troughs, with a central axis of the preprocessed vertical projection curve of the N-th frame being a symmetry axis, in the vertical projection curve of the N-th frame, as the coordinates of the left and right eyeballs in the horizontal direction of the N-th frame; wherein the vertical projection curve of the N-th frame is preprocessed by selecting peak values with their distances between a trough and a peak being above the second threshold in the vertical projection curve of the N-th frame.
This invention relates to a method for determining the horizontal coordinates of left and right eyeballs in a video frame using vertical projection analysis. The method addresses the challenge of accurately locating eye positions in images, which is critical for applications like gaze tracking, facial recognition, and human-computer interaction. The process involves analyzing the vertical projection curve of a frame, which represents the sum of pixel intensities along vertical lines. The curve is preprocessed to filter out noise by retaining only peak values where the distance between a trough and a peak exceeds a predefined second threshold. The preprocessed curve is then analyzed to identify two symmetric troughs, with the central axis of the curve serving as the symmetry reference. The horizontal coordinates of these troughs correspond to the positions of the left and right eyeballs in the frame. This approach enhances accuracy by leveraging symmetry and filtering irrelevant data, ensuring reliable eye detection even in varying lighting conditions or partial occlusions. The method is particularly useful in real-time systems where precise eye tracking is required.
7. The method according to claim 3 , wherein determining the central position of the pupil according to the coordinates of the left and right eyeballs in the vertical direction and horizontal directions of the N-th frame comprises: selecting pixels having grayscale values below a third threshold in an area defined by the coordinates of the left and right eyeballs in the vertical and horizontal directions of the N-th frame to constitute two sets of positions of pupils of the left and right eyeballs, wherein the two sets of positions consist of coordinates of the pixels having grayscale values below the third threshold in the area; and determining a centroid of one of the two sets of positions as the central position of the pupil.
This invention relates to pupil tracking in eye imaging systems, specifically improving the accuracy of determining the central position of a pupil from captured eye images. The problem addressed is the difficulty in precisely locating the pupil center due to variations in lighting, eye movement, and image noise, which can lead to inaccurate gaze tracking or biometric authentication. The method involves analyzing an image frame (N-th frame) containing left and right eyeballs. First, an area around the coordinates of the left and right eyeballs is defined in both vertical and horizontal directions. Within this area, pixels with grayscale values below a predefined threshold (third threshold) are selected. These pixels form two sets of positions, each representing potential pupil locations for the left and right eyeballs. The coordinates of these pixels are used to calculate the centroid of each set, which is then designated as the central position of the respective pupil. This approach enhances accuracy by focusing on low-grayscale regions, which typically correspond to the darker pupil area, while filtering out irrelevant data. The method is particularly useful in applications requiring precise eye tracking, such as virtual reality, medical diagnostics, or security systems.
8. The method according to claim 1 , wherein determining the area of the face grayscale image corresponding to the eyeball window comprises: making leftward and rightward extensions over a first preset distance in a horizontal direction of the N-th frame, and making upward and downward extensions over a second preset distance in a vertical direction of the N-th frame, wherein each of the extensions has the central position of the pupil as a center; and determining areas extended over the N-th frame with the central position of the pupil being the center of the extensions as the area corresponding to the eyeball window.
This invention relates to image processing techniques for analyzing facial images, specifically focusing on the detection and isolation of the eyeball region within a grayscale image of a face. The problem addressed is the accurate identification of the eyeball area in a facial image, which is crucial for applications such as gaze tracking, eye movement analysis, and biometric authentication. The method involves processing a sequence of image frames, each containing a grayscale representation of a face. For a given frame (N-th frame), the central position of the pupil is first identified. Using this central position as a reference point, the method extends horizontally (leftward and rightward) over a predefined distance (first preset distance) and vertically (upward and downward) over another predefined distance (second preset distance). These extensions define the boundaries of the eyeball window, which is the region of interest for further analysis. The resulting area, centered on the pupil, is designated as the eyeball window area in the frame. This approach ensures that the eyeball region is consistently and accurately isolated, even if the pupil's position varies between frames. The predefined extension distances allow for flexibility in adapting to different facial structures and imaging conditions, while maintaining precision in the localization of the eyeball. This technique is particularly useful in applications requiring real-time or high-accuracy eye tracking.
9. The method according to claim 1 , wherein determining the contour of the iris in the N-th frame according to the area corresponding to the eyeball window comprises: starting to search from the central position of the pupil and within the area corresponding to the eyeball window according to a preset condition; determining a grayscale value of a position to which a search is made as a reference grayscale value of the search; if a difference between a reference grayscale value of an M-th search and a reference grayscale value of an (M+1)-th search is above a fourth threshold, then determining a point to which the M-th search is made as a point at an edge of the iris, wherein M is a positive integer; and determining the contour of the iris in the N-th frame according to points at edges of the iris.
This invention relates to iris contour detection in digital images, particularly for applications in biometric authentication or eye-tracking systems. The problem addressed is accurately identifying the boundary of the iris within an image frame, which is challenging due to variations in lighting, pupil dilation, and image noise. The method involves analyzing a sequence of image frames to determine the iris contour in each frame. For the N-th frame, the process begins by defining an eyeball window area that includes the iris and pupil. A search is initiated from the center of the pupil and proceeds outward within this window. At each search position, the grayscale value is recorded as a reference value. The algorithm compares consecutive search positions (M-th and (M+1)-th) and checks if the difference in their grayscale values exceeds a predefined threshold. If it does, the position of the M-th search is marked as an edge point of the iris. This process repeats across the window to identify multiple edge points, which are then used to reconstruct the full contour of the iris in the N-th frame. The method ensures robustness by dynamically adjusting to local grayscale variations, improving accuracy in iris boundary detection.
10. The method according to claim 9 , wherein the preset condition is that each search is made over a distance d at a searching angle of a1+(x−1)λ; a1 is a first angle threshold, x is the number of searches, and λ is a second angle threshold; the method further comprises: when the searching angle is more than or equal to a2, stopping searching within the area corresponding to the eyeball window; a2 is a third angle threshold, a2 is more than a1, and a2 is more than λ.
This invention relates to an eye-tracking method for determining the position of an eyeball within a defined area, such as an eyeball window. The problem addressed is improving the efficiency and accuracy of eyeball position detection by optimizing the search process within the defined area. The method involves performing multiple searches over a distance d at progressively adjusted angles. Each search is conducted at an angle of a1 + (x−1)λ, where a1 is a predefined initial angle threshold, x is the search iteration number, and λ is a second angle threshold that determines the angular increment between searches. This iterative approach ensures systematic coverage of potential eyeball positions. The method also includes a stopping condition: if the searching angle reaches or exceeds a third angle threshold a2 (where a2 is greater than both a1 and λ), the search process terminates within the eyeball window. This prevents unnecessary searches beyond a reasonable range, improving computational efficiency. The method is particularly useful in applications requiring precise and rapid eye-tracking, such as gaze-based interfaces or medical diagnostics.
11. An apparatus for eye-tracking, comprising: a processor; and a memory storing at least one instruction, wherein the processor is configured to execute the at least one instruction to: obtain a set of face grayscale images to be detected, wherein the set of face grayscale images comprises a plurality of frames of face grayscale images; judge whether a contour of an iris is determined in an N-th frame of the plurality of frames, wherein N is a positive integer; detect a pupil in the N-th frame and determine a central position of the pupil in the N-th frame when the contour of the iris is not determined in the N-th frame; determine an area of the N-th frame corresponding to an eyeball window, wherein the central position of the pupil is a center of the area; and determine the contour of the iris in the N-th frame according to the area corresponding to the eyeball window.
This invention relates to an eye-tracking apparatus designed to improve the accuracy of iris contour detection in grayscale face images. The system addresses challenges in real-time eye-tracking where iris contours may be difficult to detect due to lighting conditions, occlusions, or motion blur. The apparatus includes a processor and memory storing instructions for processing a sequence of grayscale face images (frames). For each frame, the system first checks if the iris contour is already determined. If not, it detects the pupil and calculates its central position. Using this position, the system defines an eyeball window—a localized region of interest centered on the pupil. Within this window, the apparatus then refines the detection of the iris contour. This approach enhances tracking reliability by focusing computational resources on the relevant area, reducing errors from background noise or irrelevant facial features. The method dynamically adapts to each frame, ensuring robust performance even in suboptimal conditions. The system is particularly useful in applications requiring precise gaze estimation, such as augmented reality, medical diagnostics, or human-computer interaction.
12. The apparatus according to claim 11 , wherein when the contour of the iris is determined in the N-th frame, then the processor is further configured to execute the at least one instruction to: make the central position of the pupil determined according to the contour of the iris in the N-th frame as a central position of the pupil in an (N+1)-th frame; determine an area of the (N+1)-th frame corresponding to an eyeball window, wherein the central position of the pupil in the (N+1)-th frame is a center of the area of the (N+1)-th frame corresponding to the eyeball window; and determine a contour of the iris in the (N+1)-th frame according to a preset condition and the area of the (N+1)-th frame corresponding to the eyeball window.
This invention relates to iris recognition systems, specifically improving the accuracy and efficiency of pupil and iris contour detection in sequential image frames. The problem addressed is the computational complexity and potential inaccuracies in real-time iris recognition when processing multiple frames, particularly in tracking the pupil and iris contours across consecutive frames. The apparatus includes a processor configured to analyze image frames of an eye. When the contour of the iris is determined in the N-th frame, the processor uses the central position of the pupil from that frame as the initial central position for the pupil in the subsequent (N+1)-th frame. The processor then defines an eyeball window area in the (N+1)-th frame, centered on this pupil position. Within this window, the processor determines the iris contour based on preset conditions, such as edge detection or pattern matching algorithms. This approach reduces computational overhead by limiting the search area for iris contour detection in each new frame, leveraging the spatial consistency of the pupil position between consecutive frames. The method ensures more reliable tracking of the iris in dynamic conditions, such as slight eye movements or changes in lighting.
13. The apparatus according to claim 11 , wherein the processor is further configured to execute the at least one instruction to: perform horizontal integral projection for the N-th frame, and obtain a horizontal projection curve of the N-th frame; determine coordinates of left and right eyeballs in a vertical direction of the N-th frame according to the horizontal projection curve of the N-th frame; perform vertical integral projection for an area of the N-th frame corresponding to coordinates of the left and right eyeballs in the vertical direction of the N-th frame, and obtain a vertical projection curve of the N-th frame; determine coordinates of the left and right eyeballs in a horizontal direction of the N-th frame according to the vertical projection curve of the N-th frame; and determine the central position of the pupil according to the coordinates of the left and right eyeballs in the vertical and horizontal directions of the N-th frame.
This invention relates to a system for detecting pupil positions in video frames, addressing challenges in accurately tracking eye movements for applications like gaze estimation or user interaction. The system processes video frames to determine the central position of pupils by analyzing horizontal and vertical projection curves. For a given frame, the system first performs a horizontal integral projection to generate a horizontal projection curve, which helps identify the vertical coordinates of the left and right eyeballs. Next, it performs a vertical integral projection on the region corresponding to these vertical coordinates to obtain a vertical projection curve, which refines the horizontal coordinates of the eyeballs. By combining these vertical and horizontal coordinates, the system calculates the central position of the pupil. This method enhances accuracy in pupil detection by leveraging projection-based analysis in both directions, reducing errors from noise or partial occlusions. The system is particularly useful in real-time applications requiring precise eye-tracking, such as augmented reality, medical diagnostics, or human-computer interaction.
14. The apparatus according to claim 13 , wherein the processor is further configured to execute the at least one instruction to: remove pixels having grayscale values above a first threshold in the N-th frame, then perform horizontal integral projection for the N-th frame, and obtain the horizontal projection curve of the N-th frame.
This invention relates to image processing techniques for analyzing grayscale images, particularly in applications requiring frame-by-frame analysis such as video processing or machine vision. The problem addressed is the need to efficiently extract meaningful data from grayscale images by reducing noise and enhancing relevant features before further analysis. The apparatus includes a processor configured to process a sequence of grayscale image frames. For a given N-th frame, the processor first removes pixels with grayscale values exceeding a first threshold, effectively filtering out bright or saturated regions that may obscure relevant features. After thresholding, the processor performs a horizontal integral projection on the remaining pixels, summing pixel values along each horizontal line to generate a horizontal projection curve. This curve represents the distribution of pixel intensities across the frame's height, providing a compact representation of the frame's horizontal structure. The horizontal projection curve can be used for various purposes, such as detecting edges, identifying regions of interest, or tracking objects across frames. By applying thresholding before projection, the method reduces the influence of irrelevant bright regions, improving the accuracy of subsequent analyses. This approach is particularly useful in applications where real-time processing is required, such as surveillance, industrial inspection, or autonomous navigation. The technique ensures that only relevant pixel data contributes to the projection, enhancing the robustness of the analysis.
15. The apparatus according to claim 13 , wherein the processor is further configured to execute the at least one instruction to: preprocess the horizontal projection curve of the N-th frame, and determine coordinates in the N-th frame corresponding to an area between a second trough and a third trough of the preprocessed horizontal projection curve of the N-th frame as the coordinates of the left and right eyeballs in the vertical direction of the N-th frame; wherein the horizontal projection curve of the N-th frame is preprocessed by selecting values of peaks in the horizontal projection curve, and a distance of each of the peaks along the vertical direction of the N-th frame from a nearest trough or crest of the horizontal projection curve is greater than a second threshold.
This invention relates to image processing for eye detection in video frames, specifically improving the accuracy of identifying eyeball positions using horizontal projection curves. The problem addressed is the difficulty in reliably detecting eyeball coordinates in the vertical direction due to noise and variations in facial features. The solution involves preprocessing a horizontal projection curve of a video frame to enhance peak detection, particularly for identifying troughs that correspond to eyeball positions. The processor analyzes the horizontal projection curve of the N-th frame, selecting peaks where the vertical distance to the nearest trough or crest exceeds a predefined threshold. This filtering step ensures only significant peaks are considered, reducing false positives. The coordinates between the second and third troughs of the preprocessed curve are then determined as the vertical positions of the left and right eyeballs. This method improves robustness by focusing on distinct features in the projection curve, making it less sensitive to minor variations in facial structure or lighting conditions. The preprocessing step ensures that only meaningful peaks are used, enhancing the accuracy of eyeball localization in video frames.
16. The apparatus according to claim 13 , wherein the processor is further configured to execute the at least one instruction to: preprocess the vertical projection curve of the N-th frame, and determine coordinates, in the N-th frame, corresponding to two symmetric troughs, with a central axis of the preprocessed vertical projection curve of the N-th frame being a symmetry axis, in the vertical projection curve of the N-th frame, as the coordinates of the left and right eyeballs in the horizontal direction of the N-th frame; wherein the vertical projection curve of the N-th frame is preprocessed by selecting peak values with their distances between a trough and a peak being above the second threshold in the vertical projection curve of the N-th frame.
This invention relates to image processing for detecting eye positions in video frames. The problem addressed is accurately identifying the horizontal coordinates of left and right eyeballs in a frame by analyzing vertical projection curves. The apparatus includes a processor configured to preprocess a vertical projection curve of an N-th frame to filter relevant peak values. Specifically, the processor selects peak values where the distance between a trough and a peak exceeds a second threshold, effectively removing noise or irrelevant features. The preprocessed curve is then analyzed to determine the coordinates of two symmetric troughs, with the central axis of the curve serving as the symmetry axis. These trough coordinates correspond to the horizontal positions of the left and right eyeballs in the frame. The method ensures robust eye detection by focusing on significant features in the projection curve, improving accuracy in applications like gaze tracking or facial recognition. The preprocessing step enhances reliability by filtering out minor fluctuations, ensuring only meaningful peaks are considered for eye position determination.
17. The apparatus according to claim 13 , wherein the processor is further configured to execute the at least one instruction to: select pixels having grayscale values below a third threshold in an area defined by the coordinates of the left and right eyeballs in the vertical and horizontal directions of the N-th frame to constitute two sets of positions of pupils of the left and right eyeballs, wherein the two sets of positions consist of coordinates of the pixels having grayscale values below the third threshold in the area; and determine a centroid of one of the two sets of positions as the central position of the pupil.
This invention relates to eye-tracking technology, specifically improving pupil detection in video frames. The problem addressed is accurately identifying pupil positions in images where lighting conditions or noise may obscure the pupils, leading to errors in gaze tracking. The apparatus includes a processor configured to analyze video frames to detect and track eye movements. For a given frame (N-th frame), the processor identifies an area around the coordinates of the left and right eyeballs. Within this area, pixels with grayscale values below a predefined threshold (third threshold) are selected. These pixels form two sets of coordinates, each representing the potential positions of the left and right pupils. The processor then calculates the centroid of each set, using this centroid as the central position of the pupil. This method enhances accuracy by focusing on low-grayscale regions, which typically correspond to the dark pupils, while filtering out irrelevant data. The invention builds on prior eye-tracking techniques by refining pupil localization through threshold-based pixel selection and centroid calculation, improving robustness in varying lighting conditions. The approach ensures reliable pupil detection even when partial occlusion or noise affects the image.
18. The apparatus according to claim 11 , wherein the processor is further configured to execute the at least one instruction to: make leftward and rightward extensions over a first preset distance in a horizontal direction of the N-th frame, and make upward and downward extensions over a second preset distance in a vertical direction of the N-th frame, wherein each of the extensions has the central position of the pupil as a center; and determine areas extended over the N-th frame with the central position of the pupil being the center of the extensions as the area corresponding to the eyeball window.
This invention relates to image processing for eye tracking, specifically determining an eyeball window area in a video frame. The problem addressed is accurately identifying the region of interest around a pupil in an image to facilitate precise eye tracking. The apparatus includes a processor that processes video frames to track pupil movement. The processor extends a rectangular or square window around the pupil's central position in the N-th frame. The extensions are made horizontally (left and right) over a first preset distance and vertically (up and down) over a second preset distance, with the pupil's center as the reference point. The resulting extended area is defined as the eyeball window, which dynamically adjusts based on the pupil's position. This method ensures the window captures the entire eyeball region, improving tracking accuracy. The preset distances can be adjusted to optimize performance based on factors like image resolution or pupil size. The invention enhances eye-tracking systems by providing a reliable method to define the tracking window around the pupil.
19. The apparatus according to claim 11 , wherein the processor is further configured to execute the at least one instruction to: start to search from the central position of the pupil and within the area corresponding to the eyeball window according to a preset condition; determine a grayscale value of a position to which a search is made as a reference grayscale value of the search; if a difference between a reference grayscale value of an M-th search and a reference grayscale value of an (M+1)-th search is above a fourth threshold, then determine a point to which the M-th search is made as a point at an edge of the iris, wherein M is a positive integer; and determine the contour of the iris in the N-th frame according to points at edges of the iris.
This invention relates to image processing for eye tracking, specifically detecting the iris contour in video frames. The problem addressed is accurately identifying the iris boundary in real-time eye tracking systems, where variations in lighting and pupil movement can lead to detection errors. The apparatus includes a processor configured to analyze video frames of an eye. The processor first locates the pupil's central position and defines an eyeball window area around it. It then searches for the iris edge starting from the pupil center, using a preset condition to guide the search. During the search, the processor records grayscale values at each search position as reference values. If the grayscale difference between consecutive searches (M-th and (M+1)-th) exceeds a predefined threshold, the position from the M-th search is marked as an iris edge point. This process repeats across the eyeball window to collect multiple edge points, which are then used to determine the iris contour in the current frame. The method ensures robust iris detection by dynamically adjusting to grayscale variations, improving accuracy in eye tracking applications. The system is designed for real-time processing, making it suitable for gaze tracking in augmented reality, medical diagnostics, or user interface control.
20. The apparatus according to claim 19 , wherein the preset condition is that each search is made over a distance d at a searching angle of a1+(x−1)λ; a1 is a first angle threshold, x is the number of searches, and λ is a second angle threshold; the processor is further configured to execute the at least one instruction to stop searching within the area corresponding to the eyeball window when the searching angle is more than or equal to a2; a2 is a third angle threshold, a2 is more than a1, and a2 is more than λ.
This invention relates to an apparatus for tracking eye movements, specifically addressing the challenge of accurately determining the position of an eyeball within a defined window area. The apparatus includes a processor configured to execute instructions for performing multiple searches within the window area to locate the eyeball. Each search is conducted over a distance d at a searching angle of a1+(x−1)λ, where a1 is a predefined initial angle threshold, x is the number of searches performed, and λ is a predefined angular increment. The processor stops the search process when the searching angle reaches or exceeds a second predefined angle threshold a2, which is greater than both a1 and λ. This method ensures efficient and precise tracking by systematically adjusting the search angle based on the number of attempts, while preventing unnecessary searches beyond a maximum threshold angle. The apparatus optimizes eye-tracking accuracy by dynamically adapting the search parameters to the eyeball's position within the defined window area.
21. A non-transitory computer readable storage medium, storing computer instructions configured to enable a computer to perform the method for eye-tracking according to claim 1 .
This invention relates to eye-tracking technology, specifically a method for tracking eye movements using computer vision techniques. The system captures images of a user's eyes using a camera, processes these images to detect and analyze eye features such as pupils and iris boundaries, and calculates the gaze direction based on these features. The method includes steps for image preprocessing, feature extraction, and gaze estimation, which may involve machine learning models trained on eye-tracking data. The system may also compensate for head movements by tracking facial landmarks and adjusting gaze calculations accordingly. The stored instructions enable a computer to execute this method, providing real-time or post-processing eye-tracking capabilities for applications such as human-computer interaction, virtual reality, or accessibility tools. The invention improves upon existing eye-tracking systems by enhancing accuracy, reducing computational overhead, or adapting to varying lighting conditions. The non-transitory storage medium ensures the instructions are persistently available for execution.
22. A computer program product, comprising a computer program stored in a non-transitory computer readable storage medium, wherein the computer program comprises program instructions, and when the program instructions are executed by a computer, the computer is configured to perform the method for eye-tracking according to claim 1 .
Eye-tracking technology is used to monitor and analyze the movement of a user's gaze, often for applications in human-computer interaction, accessibility, and research. A challenge in eye-tracking systems is accurately detecting and tracking eye movements in real-time while minimizing computational overhead and ensuring robustness across varying lighting conditions and user positions. This invention addresses these challenges by providing a computer program product stored on a non-transitory computer-readable storage medium. The program includes instructions that, when executed by a computer, enable the system to perform eye-tracking by capturing images of a user's eyes using a camera, processing the images to detect and track eye movements, and analyzing the tracked data to determine gaze direction or fixation points. The system may employ machine learning or image processing techniques to enhance accuracy and efficiency. Additionally, the program may include calibration routines to adapt to different users or environmental conditions, ensuring consistent performance. The solution aims to provide a reliable, low-latency eye-tracking method suitable for integration into devices such as smartphones, VR headsets, or assistive technologies.
Unknown
June 30, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.